datacleanr 0.1.2

Clean messy real-world datasets with safe, explainable defaults before ML.

datacleaner

Clean messy real-world datasets before machine learning using safe, explainable defaults.

Installation

pip install datacleanr

PyPI: https://pypi.org/project/datacleanr/

Quick Start

from datacleaner import clean

df = clean(df)

Key Features

• Missing value handling with high-missing column drop and median/mode fill
• Duplicate row removal
• Datatype fixing with safe numeric and datetime conversion
• Outlier handling with IQR-based cap or remove modes
• Text standardization for common inconsistencies
• Feature selection to drop useless columns
• Correlation-based feature reduction
• Dataset analysis before cleaning
• Safety mode to prevent excessive data loss through step rollback

Example Usage

from datacleaner import analyze, clean

analysis = analyze(df)
df_clean, report = clean(df, return_report=True, verbose=True)

print(df.shape, "->", df_clean.shape)
print(report["summary"]["actions_summary"])

This workflow runs a quick pre-cleaning analysis, applies the default cleaning pipeline, and returns a structured report for review.

Why This Library Exists

Messy datasets routinely slow down model development. Teams often repeat similar preprocessing work across projects, and ad-hoc cleaning scripts make behavior inconsistent. datacleaner provides a consistent baseline pipeline with transparent diagnostics so cleaning decisions are easier to review and reuse.

Important Note

• PyPI package name: datacleanr
• Python import name: datacleaner

Install with pip using datacleanr, then import in code using datacleaner.

Project Structure

datacleaner/
├── src/datacleaner/
├── examples/
├── README.md
├── pyproject.toml
├── LICENSE
└── .gitignore

Core modules in src/datacleaner include cleaning orchestration, analysis, missing values, duplicates, datatypes, outliers, text standardization, column selection, correlation reduction, and reporting.

Contributing

Pull requests are welcome. Open an issue to discuss bugs, improvements, or proposed changes before large updates.

License

MIT License. See LICENSE.